A busy border inside a cell
A nuclear pore is a passage for molecular cargo between the nucleus and the surrounding cell. It is built from many protein components. Understanding their arrangement helps scientists investigate how this transport machinery works. [3]
The puzzle: where do the pieces fit?
In 2022, researchers studied the outward-facing ring of a pore from African clawed frog egg cells. Microscope measurements revealed an overall map, but fitting the protein components into it was difficult. [3] Think of assembling a complicated structure when you can see its outline but cannot clearly distinguish every piece. That is an analogy, not a description of the experiment’s equipment.
Reconstructed density map
Candidate protein structures
Fit components to observed density
Simplified workflow from the nuclear-pore study in source 3.
What AI contributed
AlphaFold supplied predicted protein shapes. The team fitted those shapes into the microscope-derived map and used further predictions to investigate interactions. Together, the approaches produced a nearly complete model of this ring. [3]
What scientists learned
The model identified arrangements of components, including five copies of a protein called Nup358. The achievement was a more detailed view of part of a molecular transport machine—not a treatment or a complete map of every nuclear pore. [3]
Why this is a discovery story
The interesting combination is a physical observation and a computational proposal working together. One supplied evidence about the assembly; the other helped interpret its pieces. Readers who want to know how researchers judge those predictions can explore the confidence measures below.
Go a little deeper
Optional reading · about 1 more minute
How confident is a predicted shape?
AlphaFold training distinguishes local confidence (pLDDT) from predicted aligned error (PAE), which helps interpret relative positions. A well-predicted piece can still have an uncertain placement relative to another piece. These computational measures help interpret a model; they are not experimental confirmation. [1] [2]
Original sources
Attributed synthesis, not original reporting. Examples labeled hypothetical or illustrative are explanatory. Reviewing a source does not independently validate its findings.
- EMBL-EBI: AlphaFold2 inputs and outputs ↗
Training resource; page publication date not established.
- EMBL-EBI: Evaluating predicted structures ↗
Training developed with Google DeepMind; not independent replication.
- Research paper: Integrative cryo-EM and AlphaFold ↗
Published in Science, June 10, 2022; abstract and methods context reviewed.
